Graphics processing units (GPUs) are managing increasingly complex computations which, in turn, drives increasingly large memory bandwidth requirements. There are few, if any, easy and low-cost techniques to scale up the bandwidth. Lossless compression of a rendering target stream as it is evicted from the render cache can save write bandwidth as well as read bandwidth by a sampler in case of dynamic textures or when blending is used in the rendering, or when actually displaying using the Display Buffer.